Java Solaris 10 Zone settings

UNIX Solaris
1

I have the following situation

1) Start JBoss Server 1
2) Within application - print as normal
3) Start JBoss Server 2

4) Printing STOPS on Server 1; Printing does NOT WORK on Server 2

Only solution is to stop Server 2;
Restart Server 1
Then Printing works as normal.

Doung a truss monitor (logging on jboss pid) gives error:

$ truss -f -d -p 5759 2>&1 | egrep -i 'exec|mem|fork'

A Good TRUSS output where printing works gives:

It works fine on our non-zone server which runs Solaris 10.

Any advice on Solaris/Java settings I should be looking at?

Thanks in advance

2

so server 1 is your global zone and server 2 is your non-global zone? are you inheriting all default filesystems from the global zone? my guess here is you are trying to write to a read-only fs. by default inherit-pkg-dir use the following filesystems as read-only loopback fs:

  • /lib
  • /platform
  • /sbin
  • /usr

i also saw something like this with oracle on a zone. the option was to create a whole root zone.

3

Sorry for the confusion with terminology, but, I am running one non-global zone with JBOSS App Server running as 2 separate instances (server 1 and server2) on the same Solaris server.

There are other zones on the server but for this application runs on one zone (accessing a database on another zone).

4

If your application is failing on ENOMEM, you need either to add memory (real or virtual) to your system or reduce your application requirements.

5

good point jiliagre. i found this as an interesting read and maybe something to get you into finding a solution: Mandalika's scratchpad: Siebel on Sun Solaris: Know the process resource limits

6

Hi Guys

My point is that the application works correctly in a NON-ZONED Solaris 10 server with 2GB, but in a zoned environment I am getting the failure as stated (truss ENOMEM).

Are there any specific commands/settings I should be looking for at the Solaris level?

I start my JBoss instances with java settings:

 
$JAVA_HOME/bin/java \
-server \
-Djava.endorsed.dirs="$JBOSS_HOME/lib/endorsed" \
-Dsun.rmi.dgc.client.gcInterval=300000 \
-Dsun.rmi.dgc.server.gcInterval=300000 \
-verbose:gc \
-XX:PermSize=128m \
-XX:MaxPermSize=128m \
-XX:NewRatio=4 \
-XX:+PrintGCDetails \
-XX:+PrintGCTimeStamps \
-XX:+UseParallelGC \
-XX:ThreadStackSize=512 \
-Xms384m \
-Xmx768m \
 org.jboss.Main -b0.0.0.0 -c foo "$@"

Thanks for your help.

7

Either your zone has its swap size capped and you need to increase that size or you have just not enough swap. You can increase its size easily by adding a swap file.

8

As jlliagre mentioned, it is most likely a memory cap or just a memory issue. What's the physical memory/swap on this box and what's the output of:

#zonecfg -z <zoneName> info

If the info looks good (no cap on phys/swap) then check the memory usage from the global while you're starting up server2 to see if it's hitting the ceiling. Also try to just start server2 without having server1. Did you compare the max heap sizes of both servers?

9

Solved

Thanks for the 'simple' advise. Swap was indeed the problem, we upped it to twice the RAM size and I can now run 3 JBoss instances on the same Solaris 10 server.

Thanks jlliagre